Radio Frequency Photonic In-Phase and Quadrature-Phase Vector Modulation

Transcription

1 Radio Frequency Photonic In-Phase and Quadrature-Phase Vector Modulation A Thesis Presented to The Academic Faculty By Kyle Davis In Partial Fulfillment Of the Requirements for the Degree Master of Science in Electrical Engineering Georgia Institute of Technology December, 2013 Copyright c 2013 by Kyle Davis

2 Radio Frequency Photonic In-Phase and Quadrature-Phase Vector Modulation Approved by: Dr. Stephen E. Ralph, Advisor School of Electrical and Computer Engineering Georgia Institute of Technology Dr. William D. Hunt School of Electrical and Computer Engineering Georgia Institute of Technology Dr. Benjamin D. B. Klein School of Electrical and Computer Engineering Georgia Institute of Technology Date Approved: November 12, 2013

3 ACKNOWLEDGEMENTS The author would like to thank the committee members for many fruitful discussions and for their guidance throughout the preparation of this thesis. The author would also like to thank the Institute for the facilities and personnel that helped produce the results of this document. i

4 TABLE OF CONTENTS ACKNOWLEDGEMENTS LIST OF FIGURES SUMMARY i iv vii 1 INTRODUCTION 1 2 I/Q VECTOR MODULATION 2 3 ARCHITECTURE COMPONENTS Mach-Zehnder Modulator Variable Optical Attenuators RF Hybrid Coupler Alternatives RF Circuit Design Optical Carrier Suppressed Single Side Band (CS-SSB) PREVIOUS RF PHOTONIC I/Q DESIGNS Switch Network with RF Hybrid Theory and Design Results and Discussion RF Hybrid with Two MZMs Theory and Design Results and Discussion Carrier Suppressed Single Sideband Theory and Design Results and Discussion POTENTIAL RF PHOTONIC I/Q DESIGNS AND SIMULATIONS Basic Two VOAs and Quadrant Selection ii

5 5.1.1 Design Simulation Discussion MZM-180 -Select, Two VOAs, and 90 Selection Design Simulation Discussion Two Lasers and Quadrant Selection Design Simulation Discussion MZM Coefficient Control Design Simulation Discussion Optical Single Sideband Suppressed Carrier RF Hybrid Subset NOISE FIGURE Noise Figure Definition Shot Noise Dominated Relative Intensity Noise Dominated CONCLUSION 57 iii

6 LIST OF FIGURES List of Figures 1 I/Q block diagram for radar receiver [1] I/Q block diagram of mathematical operation flow Block diagram of fundamental I/Q vector modulation concept Block diagram of a RF I/Q vector modulator with amplitude control in electricaldomain Block diagram of a basic RF photonic link with optical gain control Block diagram of a RF photonic I/Q phase shifter with amplitude control in optical-domain Block diagram of a MZM Magnitude of E out,1 with V π = Magnitude of E out,1 with V π = Gain vs. frequency when biased close to linear region Gain vs. frequency when not biased close to linear region RSoft simulation setup for simple MZM tests Optical comb generation [2] Branch-line coupler for 90 hybrid GHz MMIC Phase Shifter [5] Amplitude and phase variation of 1-22GHz MMIC Phase Shifter [5] Block diagram of a RF photonic I/Q phase shifter with amplitude control in optical-domain Block diagram of a RF photonic I/Q phase shifter using switch network and dual port MZM [6] iv

7 19 Diagram of switch quadrant selection [6] Results of switch network RF photonic phase shifter at 100MHz [6] Time delay versus applied voltage of Hittite HMC910LC4B broadband analog time delay DC-24GHz Ring Resonator concept for flat group delay over selected frequencies [8] photonic I/Q Vector Modulator utilizing two MZMs [9] Photonic I/Q Vector Modulator utilizing four MZMs [9] Phase error over different phase settings Data sheet screen shot of commercially available Hittite Wideband RF phase shifter phase error Phase change over RF attenuation by optical attenuation weighting RF phase shifter design using optical carrier suppressed single sideband modulation [11] RF phase vs. control voltage of Optical Phase Modulator section [11] RF power vs. control voltage of Optical Phase Modulator section [11] port optical RF phase shifter array [11] Measured phase results from one output port [11] Measured power results from one output port [11] Block diagram of a proposed RF photonic I/Q vector modulator using two VOAs and an RF quadrant selector Diagram of switch quadrant selection [6] RSoft simulation setup for the basic two VOA design with quadrant selection GHz RSoft simulation of phase shifting for the basic two VOA design with quadrant selection v

8 38 10GHz RSoft simulation of quadrant phase shifting for the basic two VOA design with quadrant selection GHz RSoft simulation of amplitude control for the basic two VOA design with quadrant selection Block diagram of RF photonic vector modulator using modified quadrant selection RSoft simulation setup of RF photonic vector modulator using modified quadrant selection GHz quadrant selection RSoft simulation of RF photonic vector modulator using modified quadrant selection Block diagram of two lasers and quadrant selection RF photonic I/Q vector modulator RSoft simulation setup of RF photonic vector modulator using two lasers and a quadrant selector GHz RSoft simulation of phase shifting using two lasers and a quadrant selector Block diagram of two lasers and quadrant selection RF photonic I/Q vector modulator RSoft simulation setup of RF photonic vector modulator using MZM and a quadrant selector GHz RSoft simulation of phase shifting a MZM and a quadrant selector RSoft simulation setup for optical RF 90 hybrid RSoft simulation setup for simple MZM tests vi

9 SUMMARY The focus of this thesis is to investigate the implementation of Radio Frequency (RF) In- Phase and Quadrature-Phase (I/Q) vector modulation through the use of modern photonic components and sub-systems which offer extremely wide RF intrinsic bandwidths. All-electronic vector modulators suffer from frequency coverage limitations and amplitude and phase instability due to components such as phase shifters and variable gain controllers operating at or near 100% bandwidth. In stark contrast, once an RF signal has been modulated onto an optical carrier, the percent bandwidth of the RF to carrier is typically less than 0.01% percent. The fundamental mechanisms and basic electronic and photonic components needed to achieve vector modulation is introduced first. The primary electrical component required in most architectures is the 90 RF hybrid coupler, which is required to generate the RF I and Q terms. The two primary photonic building blocks, aside from the laser, electro-optic modulator and demodulator, are Mach-Zehnder Modulators (MZM) and Variable Optical Attenuators (VOA). Through the utilization of these components, multiple past architectures are explored and multiple new architectures are designed simulated. For each architecture, there is a discussion on the practical implementation. Considerations such as system complexity, integration, and sensitivity to unwanted environmental stimuli are taken into account with potential solutions to alleviate these risks. In closing, the noise figure and its impact on Spur-Free Dynamic Range (SFDR) for a basic RF photonic link is derived to provide a system-level figure of merit that can be used, in most RF applications, to determine the overall performance utility current and future designs. vii

10 1 INTRODUCTION Radio Frequency (RF) photonic processing has been a highly researched field over the past decade. The allure of using optical components for manipulating RF signals spawns from the ultra-wideband capabilities of these devices. Many traditional RF techniques such as source generation, mixing, and channelization have been implemented using optical devices. Typical telecommunication optical components operate in the spectral region around 193 THz. When an RF signal is modulated onto an optical carrier near 193 THz the percentage bandwidth of the RF to carrier is almost negligible, which translates into extremely low amplitude and phase variations over the RF bandwidth of interest. This response is in contrast to wideband electrical components which operate at, or over, 100% of their bandwidth resulting in unwanted amplitude and phase fluctuations over frequency. Next generation ultra-wideband transmission and processing systems will rely on precise amplitude and phase control of extremely high speed signals. Digital telecommunication systems will require very fast amplitude and phase modulation and demodulation while minimizing unwanted amplitude and phase fluctuations to drastically increase information density per channel. Additionally, ultra-wideband analog systems for commercial and defense markets will require components that can control a signal s complex characteristics with a high degree of fidelity and stability. The focus of this thesis is to review multiple architectures and offer new designs, to the author s knowledge, that achieve wideband vector modulation utilizing RF photonic processing. 1

11 2 I/Q VECTOR MODULATION The generation and control of a signal s In-phase and Quadrature (I/Q) components are required to implement (I/Q) vector modulation. (I/Q) generation is used extensively in analog and digital systems. Radar is one field that relies heavily on the ability to generate I and Q. Modern radars are coherent systems that require retention of the received signal s phase information after mixing down to baseband for processing. Without the I and Q channels, the baseband phase information of the received signal is ambiguous in two quadrants of the complex plane. This ambiguity can lead to false conclusions of a target s characteristics. Figure 1: I/Q block diagram for radar receiver [1] As shown in figure 1, an incoming signal is split and then mixed with the I and Q channels 2cos[Ωt] and 2sin[Ωt]. This mixing downconverts the original signal from the frequency Ω to base band. The two signals are then lowpass filtered to reject higher frequency content. The top channel is then multiplied by j. By summing the two channels, the amplitude and phase information is retained due to Euler s formula, e jx = cos(x) + jsin(x) (1) Euler s equation also shows that if the signal flow of figure 1 is reversed, the amplitude and phase of A(t)e θ(t) can be controlled by properly weighting G in each channel, which is shown 2

12 in figure 2. Figure 2: I/Q block diagram of mathematical operation flow. The output of figure 2 can be written as, A(t)e θ(t) (G 1 + jg 2 ) = A(t)e θ(t) Ce ϕ (2) where, C = G G 2 2 (3) [ ] ϕ = tan 1 G1 G 2 (4) From equation 2, it is shown that the amplitude and phase of the summed signal is dependent on the gain G 1 and G 2 resulting in a vector modulator with a phase control range of π/2. G 1 and G 2 must be set appropriately if a constant amplitude, C, is to be maintained while the phase, ϕ, is changed. The equations above represent the fundamental mathematical mechanisms for I/Q vector modulation. By splitting a signal into two channels, phase shifting one channel by π/2, controlling the amplitudes of each channel, and summing the channels, the phase and amplitude of the output signal can be controlled by the amplitudes of each of the two split channels. Figure 3 depicts a simplified block diagram of the process described above. 3

13 Figure 3: Block diagram of fundamental I/Q vector modulation concept. 4

14 3 ARCHITECTURE COMPONENTS As described previously, the mechanism for phase and amplitude control of a summed signal is the amplitude of the individual channels containing I and Q. The system is an analog process which means the phase and amplitude resolution is limited by the fidelity of control and stability of the two summed channels. Figure 4 shows a block diagram of a simplified RF vector modulator with RF amplitude control in the electrical-domain. To obtain RF amplitude control in the optical domain, the electrical signal must first be modulated onto an optical carrier. Figure 5 shows a basic RF photonic link with amplitude control. The laser is modulated with the input RF. The optical signal then enters a device that controls the amount of light transmitted. The resulting light is then detected by a photodetector which converts the RF back into an electrical signal. Figure 6 depicts the optical variation of the previous RF design for I/Q vector modulation. After the RF signal has been modulated onto an optical carrier, the signal is then optically split between two arms where amplitude control is implemented, and then detected and passed through an RF hybrid. Both systems perform the same function but have one core difference; the optical amplitude control components have a substantial advantage in bandwidth capabilities. Practical wideband electrical amplitude controllers operate at bandwidth percentages of 100% or above. The optical components operate at bandwidths on the order of THz, which makes the operational bandwidth to RF signals well under 0.01%. The importance of this property lies in the fact that the components using near to 100% or more of their bandwidth have problems with amplitude and phase variations over the operating bandwidth, resulting in an overall system error margin. The 90 hybrid coupler is used to provide the necessary phase difference and summation of the two channels for I/Q operation. Phase shifting the RF by 90 is not easily obtainable in the optical domain. There is a trade off between system complexity and bandwidth when 5

15 Figure 4: Block diagram of a RF I/Q vector modulator with amplitude control in electricaldomain. Figure 5: Block diagram of a basic RF photonic link with optical gain control. dealing with the RF hybrid. The hybrid offers a simple coupler design but has limited bandwidth. The photonic approach has extremely wide bandwidth but at the cost of increased system complexity and cost. These aspects will be further explored. The following sections will investigate the optical amplitude control and electrical phase shifting components for the proposed architectures found in later sections. Figure 6: Block diagram of a RF photonic I/Q phase shifter with amplitude control in opticaldomain. 6

16 3.1 Mach-Zehnder Modulator A Mach-Zehnder modulator (MZM) uses the principle of interferometry to vary the optical amplitude through the device. Figure 7 depicts the basic operation of a typical MZM. The incoming light is split by an optical coupler where one or both arms phase shift the channelized light before recombining with another optical coupler. The electric field scattering matrix is defined as, E out,1 E out,2 = j j 1 e jθ e jθ j j 1 E in,1 E in,2 (5) Figure 7: Block diagram of a MZM. Common high speed optical phase shifters are made from the nonlinear material LiNbO 3. The couplers can also be fabricated from the same material resulting in a very compact integrated optical modulator. The process by which the optical wave is shifted in phase through LiNbO 3 is known as Pockels effect, where the index of refraction is changed by an applied electric field across the material given by, n(e) = n rn3 E (6) n 0 is the reference refractive index, E is the electric field applied, and r is the electro-optic coefficient which is typically m. The phase of the optical wave is, V θ = n(e)k 0 L (7) 7

17 where k 0 is the wavenumber in a vacuum and L is the spatial location of the wave. Substituting equation 6 results in, θ = k 0 L(n rn3 E) (8) The electric field, E, (assumed to be a parallel plate configuration), applied across the material is given as, E = V = V d (9) Using the fact that k 0 = 2π/λ 0, the optical phase shift through the LiNbO 3 is now, θ = θ 0 π λ 0 d V rn3 (10) where θ 0 is the inherent phase shift resulting from n 0 from equation 6. The final form of the optical phase shift equation can be written as, θ = θ 0 π V V π (11) where, V π = dλ 0 Lrn 3 (12) V π is the voltage required to phase shift the optical wave by 180 with respect to the inherent phase shift θ 0, which causes either a maximum or minimum at the output of the second coupler. Because the transfer function is periodic over voltage, V π also sets the dynamic range and noise figure of the photonic system. Suffice to say, V π is extremely important when designing MZM devices. The scattering matrices from equation 5 can be collapsed to, 8

18 E out,1 E out,2 = 1 2 E in,1 (e jθ 1 e jθ 2 ) + je in,2 (e jθ 1 + e jθ 2 ) je in,1 (e jθ 1 + e jθ 2 ) E in,2 (e jθ 1 e jθ 2 ) (13) For simplicity, let the electric field amplitude entering port 2, E in,2, equal zero and let the top output be the only output considered, E out,1 = 1 2 E in,1(e jθ 1 e jθ 2 ) (14) Now let, θ 1 (V ) = θ 0 π V V π (15) θ 2 = θ 0 (16) which represents the assumption that θ 1 and θ 2 both are referenced to the same inherent phase shift but only θ 1 has a variable phase shift with respect to applied voltage. Equation 14 can now be written as, E out,1 = 1 2 E in,1e jθ 0 (e jπ V Vπ 1) (17) Figures 8 and 9 show the real part of E out,1 versus the applied voltage, V, when E in,1 and V π are set to 1 and also when E in,1 remains 1 and V π is set to 2. The two plots clearly show a sine function, which is a result of the exponential properties. Because the electric field transfer function is a sine function, curved portions exist which will limit the amplitude modulation possible. The most linear section of the transfer function is typically the ideal location to vary the electric field amplitude. Most MZM devices have an RF input and bias input. The bias input is used to control what region of the transfer function the modulation will be induced. The bias has a direct impact on spur-free dynamic range (SFDR) of an RF photonic system. 9

19 Figure 8: Magnitude of E out,1 with V π = 1. 10

20 Figure 9: Magnitude of E out,1 with V π = 2. Using the optical simulation tool RSoft, figures 10 and 11 show how bias selection impacts system performance. Figure 12 shows the simulation setup. A 10GHz signal is biased close to the linear region of the transfer function, and then biased away from the linear region. As shown, second and third harmonic spurs increase which is very problematic in many high performance analog systems. However, frequency comb generation is one application where multiple spurs are desired. Utilizing different bias regions and optical mixing properties of MZMs, highly predictable spurs can be generated [2]. The results of a comb technique are shown in figure 13. Figure 10: Gain vs. frequency when biased close to linear region. 11

21 Figure 11: Gain vs. frequency when not biased close to linear region. Figure 12: RSoft simulation setup for simple MZM tests. Figure 13: Optical comb generation [2]. 12

22 MZMs are used for their extremely fast response to applied electric fields. Modulators with bandwidths higher than 70GHz have been demonstrated [3]. The mechanism by which I/Q phase shifting operates is based on amplitude control. Therefore, with proper design, phase shifting speeds can approach update rates on the order of picoseconds if a high speed MZM is used for amplitude control. Another key benefit of Mach-Zehnder modulators is the fact that there exists an RF phase difference of 180 between the two output ports of the second 3dB coupler shown in figure 7. This unique property will be further explored in subsequent sections. 3.2 Variable Optical Attenuators Amplitude control is the central process for vector modulation in the architectures proposed thus far. The MZM achieves amplitude control through interferometry. Variable optical attenuators (VOAs) are also used for manipulating optical amplitude. Typical VOAs operate by either blocking or redirecting a percentage of the light passing through by using an electrical control signal. MEMs VOAs are among the most popular devices for optical amplitude control. The speed and attenuation range of the optical amplitude control devices are directly impact the performance of the overall photonic vector modulator system. The speed of the VOAs determines the update rate for phase and amplitude while the attenuation range, and jitter error, determine the resolution and phase shifting range possible. For example, if the two attenuation coefficients had only a maximum difference of 10dB, then the maximum phase range would be However, with an attenuation range of 20dB a phase shift of 89.4 is possible. [ ] 0dB φ = tan 1 = 84.3 (18) 10dB [ ] 0dB φ = tan 1 = 89.4 (19) 20dB 13

23 Commercial off the shelf MEMs VOAs achieve update rates of 100 s of nanoseconds with attenuation ranges covering greater than 30dB are readily available. Packaged MEMs VOAs also have a relatively small footprint making them very relevant for compact systems. MZM structures and VOAs are the two methods of amplitude control that will be used for exploration of multiple designs within this thesis RF Hybrid Coupler The RF hybrid coupler is responsible for obtaining the 90 phase difference between the two channels. As discussed earlier, the benefit of optical processing of RF signals is the extremely wide bandwidth capabilities coupled with very low variations in phase and amplitude over RF range frequencies. Because the hybrid coupler is an RF (electrical domain) component, the system bandwidth and variation are limited to the coupler s characteristics. Equation 20 gives a 90 hybrid coupler scattering matrix derived from figure 14 [4]. When designed correctly, the branch-line coupler evenly divides the RF power entering at port 1 between ports 2 and 3 while causing a 90 phase difference in the signals exiting 2 and 3. Port 4 is usually isolated, or terminated, meaning no power is coupled. It is meaningful to note the dependence on wavelength in the hybrid coupler design shown in figure 14. This has a direct impact on the operational bandwidth of the coupler. Without the use of cascaded sections of differing wavelengths, the bandwidth is limited to about 10% to 20% [4]. [ S ] = j 1 0 j j 0 1 j 0 (20) 14

24 Alternatives RF Circuit Design Figure 14: Branch-line coupler for 90 hybrid. There exist other designs for obtaining the required 90 phase difference that have the potential to improve the bandwidth and variation. One alternative consists of microstrip radial stubs to induce a broadband phase shift [5]. The circuit typically attempts to act as an all-pass filter that causes a flat phase shift over a range of frequencies. However, as with the hybrid coupler, making an RF phase shifter circuit that remains flat over frequency is not trivial. Figure 15 shows the design and photograph of a 1-22GHz MMIC Phase Shifter and figure 16 shows the amplitude and phase variations. The gain variation over all phase states appears to be in the range of about 1dB while the phase error error at 90 appears to be about 3 over frequency. 15

25 Figure 15: 1-22GHz MMIC Phase Shifter [5] Figure 16: Amplitude and phase variation of 1-22GHz MMIC Phase Shifter [5] 16

26 3.4.2 Optical Carrier Suppressed Single Side Band (CS-SSB) The hybrid coupler and RF phase shifter both suffer bandwidth limitations and amplitude and phase variations due to the devices operating near or over 100% of their respective bandwidths. Another possible solution is to handle the RF 90 phase shift in the optical domain. An optical 3dB coupler causes a phase difference of 90 between the two output ports. This optical phase difference does not translate to an RF phase difference because the optical carrier and sidebands are both phase shifted equally resulting in no overall RF phase shift when the signals beat at the photodetector. However, if either the carrier or sideband s phase was held constant while the other s changed, the RF phase would be shifted by that amount when mixed at the photodetector. This process will be further explored in subsequent sections. 17

27 4 PREVIOUS RF PHOTONIC I/Q DESIGNS This section will present RF photonic designs that provide insight into the different techniques used to achieve vector modulation. Many papers and designs focus on phase shifting the RF, thus calling their systems RF photonic phase shifters and not RF photonic vector modulators. In most cases it is assumed that amplitude control can be easily implemented into the phase shifting system, creating a vector modulator. Because of the close relationship between phase shifters and vector modulators, the two terms will be used interchangeably. Potential designs, found in section 5, will show connections to the designs of this section. 4.1 Switch Network with RF Hybrid The first RF photonic phase shifter will utilize the most basic configuration of I/Q generation and control. Figure 17 recalls the basic photonic block diagram presented earlier. As previously discussed, the optical signal carrying the RF information is split into two channels where amplitude control is induced through any number of optical amplitude controlling devices. The signals are converted back to the electrical domain via photodetectors/photoreceivers and then combined with an RF hybrid coupler causing the necessary 90 phase difference between the two channels. This design only obtains a control range of 90. The design described below will introduce a process to increase the phase shifting range to 360 using a network of switches and by profiting from the 180 properties of dual port MZM devices Theory and Design Figure 18 shows the block diagram given in the paper by Erwin H.W. Chan and Robert A. Minasian [6]. The laser source provides the necessary optical carrier, while the MZM electrooptically modulates the RF signal, V RF (t), onto the optical carrier. V bias is the voltage used to set the MZM to its most linear region. The two 3dB couplers and switches are used to select 18

28 Figure 17: Block diagram of a RF photonic I/Q phase shifter with amplitude control in opticaldomain. the quadrants to obtain the full 360 of phase shift. The VOAs are used for amplitude control of the two RF signals combined at the 90 hybrid coupler. A dual port MZM is used to obtain a 180 RF phase shift between the two optical output ports. This result is apparent when examining the MZM scattering equation below. Utilizing this fact of MZM operation, two switches and two 3dB couplers are used to select whether the inputs to the hybrid coupler are either 0 /0, 0 /180, 180 /0, or 180 /180 corresponding to selection of the four quadrants. Figure 19 shows the diagram used in the paper to describe how the switches are operated for full a 360 phase range. The a and b ratio refers to the amplitude ratio between the two arms entering the hybrid coupler. Equation 22 shows the dependence of amplitude ratio on phase of the combined signal. Figure 18: Block diagram of a RF photonic I/Q phase shifter using switch network and dual port MZM [6]. 19

29 E out,1 E out,2 = j j 1 e jθ e jθ j j 1 E in,1 E in,2 (21) Figure 19: Diagram of switch quadrant selection [6]. φ = tan 1 [ b a ] (22) Results and Discussion Figure 20 shows the simulated and experimental results obtained by the group using an 100MHz signal. The simulation and experimental products are in very good agreement across the full range. Recall the equation below, C = a 2 + b 2 (23) The group was concerned with holding the amplitude, C, constant while moving the phase. This was accomplished by selecting values which held C to a ratio of 1 while stepping through the 360. The group obtained a 0.2dB variation over the entire phase range. Recall that the earlier RF circuit design had an amplitude variation of about 1dB [5]. However, the RF circuit was designed and tested at much higher frequencies than the photonic system under discussion. It will be shown in later sections that this photonic approach is still superior in gain variation even at much higher frequencies. 20

30 Figure 20: Results of switch network RF photonic phase shifter at 100MHz [6]. This photonic switch network approach is very appealing due to its simplicity. The system s update rate is limited by the VOAs, which are approaching nanosecond speeds in commercial products. The optical switch speed is also a consideration; however, nanosecond optical switches are commercially available. Because the system is entirely analog, control and resolution is dictated by the electronic control of the VOAs and switches. The two biggest limitations in this design are RF hybrid coupler characteristics and the length matching required for all paths. As shown before, hybrid couplers have bandwidth limitations and phase and amplitude variation difficulties. The 1 22GHz RF circuit could be used to generate the 90 phase difference in one arm and then combined using a broadband RF coupler. If the RF circuit were designed to only shift the signal by 90 instead of the multiple stages described, more attention could be paid to phase and amplitude variation over frequency for the single desired phase shift. Using this optimized circuit design would potentially improve variations over frequency. Current commercial optical components would also cause the system to be much larger than many applications desire. However, the rapidly 21

31 growing field of integrated photonics has already realized all the components described above at the chip level [7]. Length matching is very important because a true time delay difference between paths will cause a periodic interference pattern with nulls at frequencies where the time delay corresponds to a 180 phase difference between paths. While this may not be of importance for narrowband systems, it can be a non-trivial problem for wideband operation. There are multiple solutions to obtain the highly accurate length matching necessary for wideband operation. The simplest way is through coarse fiber length matching and then to use a fiber line stretcher for fine tuning. A pulse generator and oscilloscope can be used to tune the leading edges of the two resulting pulses. A network analyzer can also be used to tune the lengths. Viewing the gain over frequency will show the interference pattern of the coupled signals. By tuning the fiber stretcher, the pattern will change. The gain over frequency will be flat, or maintain the intrinsic transfer function of the system, when the two arms are very closely matched. Another solution to match the two paths is with a wideband RF variable true time delay. Figure 21 shows time delay versus applied voltage for a commercial Hittite chip that can achieve a true time delay of up to 70 picoseconds in the frequency range of DC-24GHz. The chip would need to be inserted after the photodetectors and before the 90 hybrid. Also, having a chip in both arms, instead of only one, would keep the gain and variations between the arms more consistent and would allow more flexibility when searching for the correct tuning. Optical ring resonators offer tunable true-time delay in a very compact size. Ring resonators have gained much interest in the photonics community over the past decade. These devices act as all-pass filters with an overall group delay. The group delay is frequency dependent. The spectral response of the ring resonator is determined by the inverse of the round trip of the ring s overall delay caused by the ring s radius. This spectral response is bell shaped and 22

32 Figure 21: Time delay versus applied voltage of Hittite HMC910LC4B broadband analog time delay DC-24GHz. requires cascaded resonators of varying round trips to cause a flatter overall group delay for a wider range of optical frequencies. Figure 22 shows the optical resonator structure and group delay versus frequency [8]. Figure 22: Ring Resonator concept for flat group delay over selected frequencies [8]. Again, the switch network design is very appealing due to the relatively simple architecture. Phase and amplitude are controlled through optical attenuators and quadrant selective through optical switches. All of these devices can be integrated onto a photonic chip. The phase resolution is dependent by the RF hybrid coupler and by the electronic control of the VOAs. Length matching all the arms has a dramatic effect on the photonic system s bandwidth and phase range. Multiple solutions for length matching were discussed and show much promise. However, as photonic chip integration becomes more precise, highly accurate length matching can be achieved in fabrication voiding the necessity for alternative length matching techniques. 23

33 The next section will discuss an approach that also utilizes an RF hybrid but its position in the system will change. 4.2 RF Hybrid with Two MZMs Theory and Design In the last section the 90 RF hybrid was used at the output of two channelized photodetectors. The next approach shifts the RF hybrid before the electro-optic conversion at the MZM. This method creates the I/Q channels before the optical components, whereas the previous method created them after. Figure 23 shows the diagram of the system under discussion [9]. This system uses two light source inputs of either differing wavelength or the same wavelength but incoherent because the two sources are coupled into the same photodetector. If both light sources were coherent then unwanted phase and amplitude noise would appear in the output RF. Also, if the wavelength difference between the light sources isn t spaced sufficiently large, unwanted spurs will arise in the bandwidth of interest. Figure 23: photonic I/Q Vector Modulator utilizing two MZMs [9]. The system in figure 23 has the same process of I/Q phase shifting through amplitude control of two channels as discussed previously. In the previous system amplitude control was accomplished through variable optical attenuators. This system varies the lasers output powers to obtain the desired summed amplitude ratio for correct phase shift. In practice, 24

34 Figure 24: Photonic I/Q Vector Modulator utilizing four MZMs [9]. variable attenuators could be used in each laser path before entering the MZM architecture. Or, varying the current driving each laser could be used for the same function. The VOA approach may provide better results than current control of the lasers. This is because VOAs do not alter the wavelength or linewidth of an optical carrier. However, unless the laser has been specifically engineered for wavelength and linewidth stability under current modulation, undesired results in the output RF may occur such as increased phase noise. Figure 24 shows a second system designed by the same group. This system requires only one light source but the addition of two more MZMs along with a photodetector that separately detects the two coherent channels. The photodetector in this system does not mix the two optical signals like a coupler would, which means coherent mixing is not an issue. The input optical carrier is split and fed into two MZMs which are used for amplitude control of the I/Q channels. These two carrier channels are then modulated with the RF signal with another set of MZMs at a 90 RF phase difference. This section of the system is the same process as the system in figure 23. Once the signals are combined at the separately detected photodetector, the RF signal can be phase shifted over a range of 90 by changing the amplitudes in the first stage of MZMs. For full 360 range of phase shift, the MZM property of 180 bi-phase operation is utilized like the switch network system described previously. By applying the 25

35 correct combination of 0 and 180 phase shifts on the top and bottom MZMs for quadrant selection, a full 360 phase shift can be obtained Results and Discussion The experimental results obtained in this paper agree very well with the theory discussed in previous sections. The notion that phase and amplitude variation over different I/Q settings is very small relative to strictly using using RF components is verified through experimentation. The following results were obtained using the first of the two designs described in the previous section. The experiment used two 1.3µm multimode lasers that were incoherent with one another, a 3GHz MZM, and a 10GHz photodetector. The tests used a 2GHz input signal and a microwave network analyzer to measure phase and amplitude. Figure 25 depicts the phase error results, which shows an error of less than 1 over a 90 range. Most high-end commercial RF phase shifters are specified at a phase error of 10, as shown in figure 26. Figure 25: Phase error over different phase settings. Figure 26: Data sheet screen shot of commercially available Hittite Wideband RF phase shifter phase error. 26

36 Figure 27 shows the phase change over an RF attenuation range of 30dB. The phase modulation due to amplitude modulation is a very common undesirable result in RF attenuators. If a system requires the phase to remain the same over an attenuation range, such as in an active cancellation system where the phase difference of 180 must be maintained, variations in phase can either seriously degrade the system s functionality or require calibration look-up tables for each attenuation value set. Calibration tables can add a huge processing burden especially if there are multiple vector modulators in the system. Also, because each vector modulator will have to be specifically characterized and calibrated, manufacturing costs and time will significantly increase. Figure 27: Phase change over RF attenuation by optical attenuation weighting. The two RF photonic phase shifters described in this section have many of the same benefits and difficulties of the switch network system described earlier. The similar benefits are the integration potential and the amplitude control of the I/Q channels remaining in the optical domain, making the phase and amplitude variation over frequency extremely low. The first design uses control of the laser s output optical power to set the I/Q channels amplitude. The paper shows the two lasers being separated from the MZM chip. This is because the MZM in this paper was made of LiNbO 3. In recent years much work has been put into fabrications that have lasers and MZMs integrated on the same chip [10]. Taking this integration one step further, photodetectors could also be integrated with the lasers and MZMs to have both RF photonic phase shifter designs each fabricated on one chip. This approach using InP can also be used for the switch network by using MZM structures instead of VOAs and switches for I/Q 27

37 channel control and quadrant selection. As discussed earlier, integration of all sections of the device is a huge benefit due to the fact that highly precise length matching between the I/Q channels is required especially for higher frequencies. The most obvious difference in the MZM structure designs and the switch network design is that the RF hybrid coupler that creates the I/Q channels is inserted at different locations in the architectures. The switch network has the I/Q generation after the photodetectors while the MZM designs have the I/Q generation preceding the electro-optic conversion provided by the MZMs. The system performance difference between the two approaches is dependent on the electrical matching of either the photodetectors or the MZMs to the RF hybrids. System performance will be degraded, or calibration required, if the responsivity and linearity differ in the photodetectors. The quantum noise added and dark current can have an effect, but only need be taken into account if extremely resolute phase shifting is required. For most applications, quantum noise and dark current differences can be ignored especially if the photodetectors are fabricated on the same substrate. The MZMs also must be closely electrically matched for precise phase shifting. Maintaining correct biasing of both MZMs to their most linear region is an additional difficulty. The MZM designs also require length matching between the arms that transmit the RF information. All lengths without the RF, such as the lengths between the laser and the MZM structures, are not required to be matched for the system to function correctly. Therefore, the lengths between the RF hybrid-mzm interface and the MZM-photodetector interface must be matched. As discussed earlier, RF variable true time delays can be used to match the two RF/optical arms or optical ring resonators/optical variable delay lines. Because the MZM designs are already highly integrated, length matching is less strenuous unless high frequencies are required. The next design requires no length matching because it does not achieve phase shifting through I/Q generation. 28

38 4.3 Carrier Suppressed Single Sideband To this point the previous designs have all relied on I/Q generation and amplitude control to obtain the desired phase shift and resulting amplitude of the output RF signal. This design relies on optical phase shifting that results in an RF phase shift. The topic of this thesis is I/Q vector modulation, which is not the operation of the following design. However, it will become apparent in the potential design section where this process can be used for the I/Q method Theory and Design As briefly discussed in previous sections, an optical phase shifter does not cause an RF phase shift. This is because the optical phase shifter operates on nanometer wavelengths and not millimeter to meter wavelengths. However, if the phase of the optical carrier and electrooptically mixed RF information is offset, the output RF signal will be phase shifted by the offset when optical-to-electrical conversion is induced at the final photodetector stage. This process relies on the ability to separate the optical carrier from the RF modulated sideband, which can be difficult if the RF is not high frequency. If the RF is about 10GHz, commercially available optical filters can separate the carrier and sideband with relative ease. Even if the laser s frequency varies as much as 5GHz, the passband of the optical filters can be made wide enough to tolerate such instability. However, if the optical spectrum had an RF sideband near or below 5GHz and the passband was the same as before, the system would loose RF phase shifting functionality. The system described below uses carrier suppressed single sideband modulation techniques to achieve separation without the use of optical filtering [11]. Figure 28 shows the design used by the group. The optical carrier is divided through multiple stages to obtain the desired result. The top most stage provides the electro-optical modulation of the RF using two optical phase shifters in an MZM architecture. By phase modulating both arms at a phase difference of 90, a single sideband of RF information with the optical carrier results when the two arms are joined 29

39 in the optical coupler. The bottom arm of the design labelled Optical Phase Modulator is used to phase shift the split optical input. Now the optical carrier output from the Optical SSB Modulator and the output of the Optical Phase Modulator can be set to differing phases. When the two signals are mixed together at a photodetector, an RF phase shift dependent on the phase difference of the two modulator sections will result. While this architecture does produce an RF phase shift, the mixing properties of the two carriers will cause undesirable power fluctuations and nonlinear phase shifting over the control voltage range. The power fluctuations over applied voltage are a result that this device operates as an interferometric modulator. If no RF signal were input to the top section, the device would operate like a typical MZM structure. The interferometric properties of combining the top and bottom sections of the device results in a power fluctuation equal to the extinction ratio inherent of the device, which is about 15dB in RF power in this paper. Obviously, an RF power variation of 15dB over the total phase range renders this device extremely limited for practical systems requiring amplitude stability while phase shifting. To mitigate this problem the group inserted a third arm into the MZM structure. The DC Balance in figure 28 is used to tune the optical phase so that the optical carrier is cancelled, or greatly attenuated, when the output of the DC Balance section and the Optical SSB Modulator section are coupled. If the optical carriers are sufficiently cancelled, the combination of the two sections results in a carrier suppressed single sideband modulation. By using this process of optical carrier cancellation, the coupling of the signal with the bottom section will not have the interferometric properties as before. Using the DC Balance allows phase shifting without the undesired power fluctuations. Figures 29 and 30 show the phase and RF power versus the control voltage of the Optical Phase Modulator section. Both figures show three plots representing the device operation without the DC Balance arm, with the arm set for a splitting ratio of α = 1, and when the splitting ratio α = J 0 ( )/ 2 where 30

40 = πv RF /V π. The splitting ratio α is derived in the paper from the square-law of the resulting photo-current produced at the photodetector after the MZM device, which is why the Bessel function J 0 ( ) is present in the optimized setting for α. After optimizing for constant amplitude while controlling a predictable linear RF phase shift, the group obtained an architecture very suitable for practical applications. Figure 28: RF phase shifter design using optical carrier suppressed single sideband modulation [11]. Figure 29: RF phase vs. control voltage of Optical Phase Modulator section [11]. 31

41 Figure 30: RF power vs. control voltage of Optical Phase Modulator section [11] Results and Discussion The device was fabricated to demonstrate its use in a practical application, namely a phased array architecture. Figure 31 shows the schematic of the 4-port device and the fabricated waveguide structure. For simplicity of design, the splitting ratio α was set to 1, which figure 30 shows will induce an RF power change of about 4dB over the full phase shift range. Figure 32 and figure 33 show the resulting phase shift and power variation over a bias voltage of V π. The system was setup for a triangle wave spanning 50ms. The triangle voltage ramped up and down between V π and +V π. As was predicted by the group, the phase shift range achieves 360 while the power variation is about 4dB. The design described in this section is very appealing due to the highly integrated and compact package. The bandwidth of the optical sections are again extremely wideband with physical limitations not impeding until frequencies well beyond 100GHz. However, the structure described uses and RF hybrid to obtain the correct optical mixing required for device operation. Therefore the system is limited by the bandwidth and stability of the RF hybrid. A common theme in all the designs described thus far, is the limitation of a system requiring 32

42 Figure 31: 4-port optical RF phase shifter array [11]. Figure 32: Measured phase results from one output port [11]. 33

43 Figure 33: Measured power results from one output port [11]. RF components. One potential design will be explored in future sections that hopes to achieve an all optically generated RF phase shift where the bandwidth limitations reside only within optical or electro-optical components. 34

44 5 POTENTIAL RF PHOTONIC I/Q DESIGNS AND SIMULATIONS This section concentrates on potential RF photonic I/Q designs. The designs proposed are considered novel to the author s knowledge. The sections leading up to this portion have laid the fundamental mechanisms by which these designs function. Each section will include a design, simulation, and discussion subsection. Each design section will recall and reiterate the processes discussed in previous sections to refresh the information needed to understand the system s functionality. Simulations will then be presented and interpreted. Finally, discussions of the pros and cons will conclude each section. All the following simulations will use RSoft. This simulation software is used throughout many laboratories and was recommended by Professor Stephen E. Ralph of Georgia Tech for close agreement of experimental and simulation data. The main focus of the simulations in this section is phase and amplitude control, not link budget or noise figure. Noise figure will be discussed in subsequent sections. This thesis focuses on novel ways of using current devices as a system. Thus the figures of merit for these simulations will rely solely on design complexity and vector modulation. 5.1 Basic Two VOAs and Quadrant Selection The next two sections propose architectures very similar to the switched network hybrid design described in section 4.1. The proposed designs differ in that they use an RF component for switching quadrants instead of the optical switching network. The designs are meant to decrease system complexity with the trade-off of bandwidth and gain flatness over frequency due to the RF component insertion. 35